1
THE RELATIONSHIP BETWEEN GRAM-NEGATIVE COLONISATION AND BLOODSTREAM 1
INFECTIONS IN NEONATES: A SYSTEMATIC REVIEW AND META-ANALYSIS 2
Laura Folgori1*, Chiara Tersigni1,2*, Yingfen Hsia1, Christina Kortsalioudaki1, Paul Heath1, Mike 3
Sharland1, Julia Bielicki1,3 4
1Paediatric Infectious Disease Research Group, Institute for Infection and Immunity, St George's University of London, 5
London, UK 6
2 Department of Health Sciences , University of Florence, Anna Meyer Children's University Hospital , Florence , Italy 7
3Paediatric Pharmacology, University Children’s Hospital Basel, Basel, Switzerland 8
* These authors contributed equally to this manuscript 9
10
Intended category: Systematic Review 11
Corresponding author: Laura Folgori 12
Mailing address: St George's University of London, Jenner Wing, Level 2, Room 2.215E, Cranmer 13
Terrace, London, SW17 0RE, United Kingdom 14
E-mail address: [email protected] 15
Telephone number: +44 20 87254851 16
17
Running title: Gram-negative bloodstream infections in colonised babies 18
Keywords: infant, newborn; Gram-Negative Bacteria; carrier state; bacteraemia; neonatal screening 19
Abstract: 321 words 20
Manuscript: 3,327 words 21
CORE Metadata, citation and similar papers at core.ac.uk
Provided by St George's Online Research Archive
2
ABSTRACT 22
Objectives: Neonates admitted to Neonatal Intensive Care Units (NICU) are at significant risk of 23
developing bloodstream infections (BSIs). Gram-negative bacteria (GNB) both colonise and infect, but 24
the association between these entities is unclear. By conducting a systematic literature review, we 25
aimed to explore the impact of factors on the association between GN colonisation and GN-BSI at both 26
baby level and unit level. 27
Methods: We searched Medline, Embase, and Cochrane Library. Observational cohort studies 28
published after 2000 up to June 2016 reporting data on the total number of neonates (0-28 days) 29
colonised with GNB assessed by rectal/skin swab culture and the total number of neonates with GN-30
BSI (same bacteria) were included. Studies were excluded if data on skin/rectal colonisation, neonates, 31
and GNB could not been identified separately. The meta-analyses along with multivariate meta-32
regression with random-effect model were performed to investigate factors associated with the GN 33
colonisation and GN-BSI at baby-level and unit-level. 34
Results: 27 studies fulfilled our inclusion criteria, 15 for the baby-level and 12 for the unit-level 35
analysis. Study heterogeneity was high, with suboptimal overall quality of reporting assessed by the 36
STROBE-NI statement (44.8% of items adequately reported). In 1,984 colonised neonates, 157 (7.9%) 37
developed GN-BSI compared with 85 of 3,583 (2.4%) non-colonised neonates. Considerable 38
heterogeneity across studies was observed. Four factors were included in the meta-regression model: 39
Gross domestic product (GDP), pathogen, outbreak, and frequency of screening. There was no 40
statistically significant impact of these factors on GN colonisation and GN-BSI in baby level. We were 41
unable to perform the multivariate meta-regression due to the insufficient reported data for unit level. 42
Conclusions: Study limitations include the small number and the high heterogeneity of the included 43
studies. While this report shows a correlation between colonisation and BSI risk, this data currently 44
doesn’t support routinely screening for GNB. The analysis of large cohorts of colonised neonates with 45
3
clinical outcomes is still needed to define the major determinants leading from colonisation to 46
infection. 47
4
INTRODUCTION 48
Babies admitted to Neonatal Intensive Care Unit (NICU) are at high risk of developing bloodstream 49
infections (BSIs) and have been identified as a critical population for the acquisition and transmission 50
of multidrug-resistant (MDR) pathogens.[1] Among them, Gram-negative bacteria (GNB) are of highest 51
concern in the neonatal population, with a global increase in the incidence rate and very limited 52
therapeutic options.[2] MDR-GNB have been found to be responsible for an increasing number of 53
NICU outbreaks, with many implications for infection control policies and practices, and mortality 54
rates reported around 30%.[3] 55
GNB can cause both colonisation and infections. In a colonised patient, the organism is found on the 56
body but is not causing any symptoms or disease. At birth, healthy neonates have no endogenous 57
microflora which is rapidly acquired through perinatal transfer of maternal vaginal and gastrointestinal 58
flora (vertical transmission) and from environmental or human sources (horizontal transmission).[4] 59
However, sick neonates who require prolonged hospitalisation are at high risk of colonisation with 60
resistant or difficult-to-treat bacteria as a result of intense and long-term exposure to antibiotics and 61
the hospital environment.[5, 6] Some studies have shown a positive association between gut 62
overgrowth and neonatal sepsis.[7, 8] Studies conducted during hospital outbreaks are broadly 63
consistent in showing a relationship between the microorganisms causing colonisation and those 64
isolated from the blood cultures of septic neonates admitted to the same unit.[9] However, the 65
mechanisms leading from colonisation to infection are still debated. 66
Screening for colonisation is usually discussed in the context of intensive care to prevent cross-67
infections and inform strategies, such as patient cohorting.[10] However, role of active surveillance 68
for GNB in informing antimicrobial empirical treatment has not yet been fully explored and evaluated 69
in neonates. Clarifying the link between GNB colonisation and infection might have a significant impact 70
on the clinical management for hospitalised babies. If a link is demonstrated, carriage screening could 71
potentially be used to stratify patients to different antibiotic regimens and, at the same time, to select 72
5
baseline treatment options at unit-level and potentially conserve broad spectrum antibiotics. By 73
conducting a systematic literature review, we aimed to explore the impact of factors on the 74
association between GN colonisation and GN-BSI at both baby level and unit level. 75
76
METHODS 77
A review protocol is available upon request. Studies were considered eligible for inclusion if reporting 78
data on neonates aged 0-28 days (Population), rectal swab/stool culture or skin swab culture to assess 79
GN colonisation (Intervention), comparing the prevalence of GN-BSI among colonised and non-80
colonised neonates (Comparison), considering GN-BSI as clinical outcome (Outcome), in neonates 81
admitted to NICU (Setting). The search was limited to studies published after 2000. Given the advances 82
in modern neonatology, the aim was to capture publications that reflect policies and practices over 83
the last 15 years. No language restriction was applied. 84
Medline (Ovid MEDLINE(R) without Revisions 1996 to June Week 2 2016), Embase (Embase 1996 to 85
2016 Week 24), and Cochrane Library (Issue 6 of 12, June 2016) databases were systematically 86
searched on June 15, 2016 with a strategy combining MeSH and free text terms for “neonate” AND 87
“colonisation” AND “bloodstream infection”. The full strategy is available as Supplementary Material. 88
Two assessments for included studies were performed. In the first one (baby-level) inclusion criteria 89
for studies were their reporting of: 1) data on neonates aged 0-28 days, 2) the total number of babies 90
colonised with GNs assessed by rectal swab/stool culture or skin swab culture, and 3) the total number 91
of GN-colonised babies who developed a concordant (caused by the same pathogen) GN-BSI. In the 92
second assessment (unit-level), inclusion criteria were studies reporting: 1) data on neonates aged 0-93
28 days, 2) the total number of babies colonised with GNs assessed by rectal swab/stool culture or 94
skin swab culture during the study period, and 3) the total number of babies with GN-BSI in the same 95
unit during the same timeframe were considered eligible for inclusion. Studies were excluded if 96
6
reporting data on multiple colonisation sites but rectal and/or skin colonisation data could not be 97
identified; studies also including children and/or adults where neonatal data could not be clearly 98
extracted; and studies reporting data on both Gram positives and GNs if GN data could not be 99
identified separately. 100
The primary outcome was to investigate the variables with an impact on the association between GN 101
colonisation and GN-BSI at both baby-level and unit-level. 102
Data on study characteristics, demographic and clinical features of included neonates, inclusion and 103
exclusion criteria, outcome definitions, microbiological methods, and total numbers of 104
colonised/infected babies was independently extracted by two different authors (LF and CT), 105
according to pre-specified criteria. In case of disagreements, these were resolved in discussion with a 106
third author (JB). 107
This study did not receive any direct funding. 108
Quality assessment 109
To assess the quality of the included studies, the Newcastle-Ottawa scale was used (Table S1).[11] 110
Moreover, to assess the quality of reporting of the included studies, the recently published 111
Strengthening the Reporting of Observational Studies in Epidemiology for Newborn Infection (STROBE-112
NI) statement was used.[12] This checklist is an extension of the STROBE statement aiming to improve 113
scientific reporting of neonatal infection studies, with the ultimate goal to increase data utility and 114
allow meta-analytical approaches. The proportion of STROBE-NI items adequately reported was 115
calculated for each study. This review complies with the PRISMA guideline.[13] 116
Statistical analysis 117
The proportion of concordant GN-BSI in colonised babies was calculated as number of 118
infections/colonised babies. Colonisation pressure was calculated as number of colonised babies/total 119
7
NICU admissions in the study period. The proportions of colonisation and infections were calculated 120
using the crude data collected as the number of colonised or infected babies/total number of neonates 121
admitted during the study period. The two-tailed Mann-Whitney U test for two independent samples 122
was used to compare the STROBE-NI score between studies primarily designed for clinical and those 123
mainly for microbiological purpose. A p-value of less than 0.05 was considered statistically significant. 124
We performed a sub-group meta-analysis along with multivariate meta-regression.[14, 15] Study 125
characteristics extracted for sub-group and meta-regression were: 1) gross domestic product (GDP) 126
(upper-middle-income countries (UMIC), lower-middle-income countries (LMIC), high-income 127
countries (HIC)); 2) pathogen (Klebsiella spp. vs other Gram-negative pathogens); 3) screening timing 128
(once vs twice a week); 4) outbreak (study carried out during outbreak vs not during outbreak). We 129
carried out baby-level and unit-level meta-analyses separately. For baby-level, the meta-analysis was 130
conducted to produce estimated risk ratio (RR) as the measure of group difference (colonisation vs 131
non-colonisation) on the rate of infection. Due to the insufficient data reported for unit-level, we used 132
the Freeman-Tukey double arcsine transformation (arcsine square root transformation [16]) to 133
calculate the weighted proportion of overall infection rate. We performed the DerSimonian and Laird 134
random-model effect using inverse variance weight method, which takes into account the within study 135
variation and between study heterogeneity. The I2 statistic was used to describe the variation across 136
studies due to heterogeneity. We defined the level of heterogeneity as low, moderate, and high 137
correspond to I2 values of 25%, 50%, and 75%.[14] As the small number of included studies, we were 138
unable to carry out publication bias in this present study.[14] The meta-analysis and meta-regression 139
were carried out using STATA version 14.0 (StataCorp). 140
141
RESULTS 142
8
Study selection and description 143
The search identified 8,543 studies. Among them, 25 papers and 2 conference abstracts fulfilled our 144
inclusion criteria and were included in the final analysis. 5,254 studies were excluded based on the 145
title, 1,338 were rejected on abstract, and 211 were rejected on full text (Figure 1). 15 studies were 146
selected for the baby-level[4, 6, 8, 17-28] and 12 for the unit-level analysis.[9, 29-39] 18 out of 27 147
studies were carried out in high-income countries (HIC),[4, 6, 8, 18, 20, 22-24, 26-28, 30-32, 34-37] 5 148
in upper middle-income countries (UMIC),[17, 19, 21, 25, 29] and 2 in lower middle-income countries 149
(LMIC)[9, 33], according to the 2016 World Bank Classification (Table 1S).[40] 20 were carried out as 150
prospective[4, 6, 8, 9, 18, 20-22, 25-33, 35, 36, 38] and 5 as retrospective studies.[17, 19, 23, 34, 37] 151
Two papers did not provide their study design.[24, 39] 8 studies were carried out during hospital 152
outbreaks.[21, 24, 28-31, 37, 38] 153
Apart from one study,[23] all papers assessed colonisation through rectal swab or stool culture (Table 154
2S). 24 (88.9%) out of 27 studies provided information about timing and frequency of microbiological 155
screening.[4, 6, 8, 9, 17-36] In nearly half of the studies, rectal/skin swabs were performed weekly 156
through the baby’s NICU stay[4, 6, 17, 19, 20, 24-30, 32, 33] whereas in 6 studies neonates were 157
screened twice a week.[8, 21, 22, 31, 35, 36] 158
To evaluate the concordance between colonising and bloodstream isolates, 15 (55.6%) out of 27 159
studies performed genotyping analyses.[4, 9, 20-22, 24, 25, 28-32, 35, 37, 39] Twelve studies 160
genotyped the isolates by pulsed field gel electrophoresis (PFGE)[4, 9, 20, 25, 28-32, 35, 37, 39] 161
whereas 3 studies performed Polymerase Chain Reaction (PCR).[21, 22, 24] Only one study assessed 162
the genotype by sequencing the pathogens.[39] 163
Only one study assessed the cost-effectiveness of the intervention.[30] 164
Quality assessment of included studies 165
9
A huge variation was highlighted in terms of study design (prospective vs retrospective, inclusion 166
criteria, different outcomes assessed), included population (gestational age, birth weight, sample 167
size), and investigated pathogens (different strains, different resistance pattern). Overall, according to 168
the STROBE-NI checklist,[12] the included studies reported adequately a mean of 44.8% (range 8.6-169
67%) of the suggested items. A statistically significant difference was highlighted in terms of 170
compliance with the checklist between studies primarily designed for clinical and those mainly for 171
microbiological purposes (47.2% vs 32.4%, p=0.034). As summary considerations on study quality in 172
general, according to the Newcastle-Ottawa scale, all studies assessed the exposure and the outcome 173
by using secure records, and all of them selected the non-exposed cohort from the same community 174
as the exposed cohort. However, very few studies demonstrated that the outcome of interest was not 175
present at the start of the study and none of them reported a statement about proportion of patients 176
who completed the follow-up (Table 1S). 177
Baby-level analysis 178
15 studies were included in the baby-level analysis,[4, 6, 8, 17-28] 3 (20.0%; 3/15) of which were 179
carried out during NICU outbreaks.[21, 24, 28] 7 (46.7%; 7/15) studies provided information about 180
demographic characteristics of the included cohort (e.g. age at screening, birth weight or gestational 181
age) (Table 3S).[4, 6, 8, 19, 22, 25, 26] The length of follow-up was reported in 6 studies.[6, 22, 24-26, 182
28] Five studies reported the interval between colonization and onset of concordant BSI.[8, 13, 16, 21-183
22] 184
Overall, a total of 8,421 neonates were screened for rectal and/or skin colonisation. Among them, 185
1,984 (23.6%) were found to be colonised by GNB. In total, 157 colonised babies experienced a BSI 186
concordant with the colonising pathogen (7.9%). A broad variation was found among the included 187
studies in terms of prevalence of concordant GN-BSIs in colonised babies (range 0.0 – 42.8%). In those 188
studies that also reported the number of non-colonized babies who developed a GN-BSI, the 189
proportion of neonates who experienced a GN-BSI was 2.4% (85/3,583). 190
10
Only one study reported the relatedness between the genotype of colonising and invasive pairs of 191
isolates.[20] In this study, 17 out of 19 strains (89.0%) had an indistinguishable PFGE pattern. 192
Meta-analysis 193
All sub-group meta-analyses results are shown in Figure 2. The random-effects inverse variance meta-194
analysis for all sub-groups demonstrated strong evidence of heterogeneity within sub-groups, and 195
heterogeneity between sub-groups. The overall estimated RRs in within sub-groups analyses did not 196
show any differences for GDP, pathogen, and outbreak. However, when conducting separate meta-197
analysis for screening frequency , RR of GN-BSI in babies screened twice/week compared with once a 198
week was 1.24 (95CI: 1.12-1.37) in the non-colonisation group and 0.95 (95%CI: 0.94-0.97) in the 199
colonisation group. I-squared (I2) estimates of 75.5% (screening twice) and 64.2% (screening once) 200
showed a different heterogeneity to the overall meta-analysis. To further explore heterogeneity 201
between studies, we performed multivariate meta-regression analysis (Table 1). All included variables 202
in the meta-regression analysis did not show statistically significant impact on GN colonisation and 203
GN-BSI in the baby-level. 204
Unit-level analysis 205
12 studies were included for the analysis at the unit-level,[9, 29-39] 5 (41.6%) of which were carried 206
out during outbreaks in the neonatal units.[29-31, 37, 38] 207
A total of 6,363 babies were included. Among them, 1,825 neonates (28.7%) had a rectal/skin swab 208
positive for GNB (Table 2). The colonisation pressure varied widely among the selected studies, 209
ranging from 1.0%[34] to 81.8%.[9] Overall, the prevalence of GN-BSIs among neonates admitted to 210
the NICUs during the same timeframe was 8.1% (516 BSI episodes/6,363 admitted babies). The rate 211
of BSIs among the different studies ranged from 0.0 to 19.8%. 212
11
In those studies evaluating the molecular epidemiology among colonising and invasive strains, PFGE 213
analysis proved to be a very useful tool to investigate the spread and clonality of isolated pathogens, 214
especially in the context of NICU outbreaks.[9, 29-32, 35, 37, 39] 215
Meta-analysis 216
The sub-group meta-analyses results for unit-level are shown in Figure 1S. Results for all within sub-217
group analyses have shown considerable high heterogeneity. This may be due to the insufficient 218
reported data in the included studies. In addition, we were unable to perform the multivariable meta-219
regression model from the available unit-level data. 220
221
DISCUSSION 222
This systematic review included 27 studies, 15 were included in the baby-level and 12 in the unit-level 223
analysis. The quality of reporting assessed by the STROBE-NI statement’s checklist was suboptimal in 224
the great majority of the published studies, with a significant difference between those primarily 225
targeting clinical research questions and those focusing on microbiological research questions. Eight 226
studies were carried out during NICU outbreaks. A total of 14,784 babies were screened for gut or skin 227
colonisation. Among babies that were colonised, 7.9% developed a concordant BSI. The overall 228
estimated RRs within sub-groups were similar for GDP, pathogen, and outbreak. In addition, the 229
within-group I2 estimates for these factors were similar. However, the RRs of GN-BSI comparing twice 230
weekly with weekly screening were 1.24 in the non-colonisation group and 0.95 in the colonisation 231
group with different I2 estimates. To explore this further, meta-regression analyses were carried out. 232
None of these factors were statistically significant associated with GN colonisation and GN-BSI at the 233
baby-level. Only one study analysed the genotypic relatedness of colonising and invasive pairs of 234
isolates. Due to the insufficient reported data for unit-level, we were not able to further explore the 235
association of these factors and the outcome of interest in present study. 236
12
Many studies over the last decade have tried to assess the association between gastrointestinal (GI) 237
bacterial flora and the onset of invasive infection in neonates. Direct translocation of bacteria from 238
the GI tract to the bloodstream through immature or damaged bowel wall (such as in case of 239
necrotizing enterocolitis) and indirect transfer via other pathways due to immaturity of defence 240
mechanisms are some of the hypotheses that have been suggested.[7] Many factors associated with 241
the NICU stay, both environment- and patient-related, have been shown to influence the status of the 242
neonatal microbiome, therefore predisposing high-risk babies to nosocomial infections.[5] 243
Treatment with broad-spectrum antibiotics, frequently experienced by hospitalised neonates,[41] 244
leads to gut colonisation with multidrug-resistant Gram-negative bacteria (MDRGN) by selecting 245
resistant flora.[42] The GI tract provides an important reservoir for antibiotic-resistant GNB that can 246
then persist throughout the NICU stay and can be easily transmitted between patients.[43] 247
The individual-level association between colonisation and BSI we observed may actually explain their 248
ecological association at unit-level. For the unit-level analysis, we were unable to determine whether 249
colonisation preceded infection in affected babies. However, there may be an additional impact of 250
cross-infections with rapid transition from colonisation to invasive infection in the face of high 251
colonisation pressure. Recently, colonisation pressure has been identified as an independent risk 252
factor for ICU-acquired MDR-infections in adults.[44] 253
Conversely, the role of carriage screening to adjust empirical regimens in colonised patients in the 254
non-epidemic setting has not been properly explored yet. Screening may have a particularly important 255
role in NICUs, to closely monitor high-risk neonates, to inform empirical treatment when resistance 256
patterns are identified, and to set up preventive interventions, such as decolonisation and 257
decontamination, to reduce the risk of invasive infections.[45] Such potential interventions have to be 258
interpreted in the light of a recent review of the interventions to control neonatal healthcare-259
associated infection outbreaks, which showed that enhanced swab-based surveillance did not prove 260
to be effective at reducing case-fatality or outbreak duration.[46] 261
13
Our review showed the different RRs associated with the frequency of screening (once vs twice a 262
week) in the infection rate of subsequent BSI in non-colonised and colonized babies. Despite the 263
multivariate meta-regression failing to demonstrate a statistically significant finding for this factor, the 264
screening time plays an important role in the clinical practice. A strategy of continuous surveillance of 265
MDRGN colonization has been discussed extensively, both as a basis for preventing cross-infection 266
and to facilitate infection control measures.[47] However, there is no consensus on the optimal timing 267
and frequency of ongoing screening. 268
The predictive value of rectal MDRGN colonisation for subsequent MDRGN bacteraemia has been 269
assessed in a number of studies in adults, with variable findings. Due to the significant implication of 270
these highly resistant infections on healthcare costs and patients outcomes, the need to develop 271
clinical prediction algorithms to identify patients potentially colonised with such organisms (and 272
therefore candidates for screening) at hospital admission has been broadly recognised.[48] 273
At the moment, the cost-effectiveness of routine rectal screening cannot be fully elucidated. Frequent 274
delays in laboratory reporting of microbiological results and increased exposure to broad-spectrum 275
antibiotics are some of the potential limits for supporting colonisation-guided versus standard empiric 276
antibiotic treatments. Without clear evidence of a significant impact on patient outcome, the 277
implementation of routine surveillance cultures in those setting where MDRGNs are rare or endemic 278
might not be warranted. 279
This review has several limitations. Firstly, the association between GNB colonisation and GN-BSI in 280
neonates must be interpreted in the light of the small number of included studies and the high 281
heterogeneity in terms of study design, included population, and investigated pathogens. Due to the 282
low number of studies included in the meta-analysis, we were unable to assess publication bias. 283
Different pathogens have been shown to have different impacts on the risk of developing invasive 284
infections in colonised neonates, and pooling data on multiple strains could have biased the 285
results.[42] Lastly, the quality of data reporting was assessed according to the STROBE-NI statement 286
14
checklist. However, this guideline was designed to improve the reporting of observational studies on 287
the epidemiology of neonatal infections, and may not have been entirely suitable for some of the 288
studies included in this review primarily designed for microbiological purpose. However, this is the 289
only specific guidance currently available for the reporting of neonatal infections. 290
The analysis of large prospective cohorts of colonised neonates with their clinical outcomes is highly 291
relevant in order to clarify the risk factors and determinants for invasive infections. This is evident 292
from the observation that although we showed a correlation between colonisation and invasive 293
disease, the majority of colonised babies do not develop systemic invasive infection. Previously 294
published studies did not attempt to link WGS data with clinical outcome nor to ascertain the 295
relatedness between colonising and invasive pathogens. Such information could assist in gaining 296
evidence on pathogenicity determinants and might have a significant impact on the management of 297
neonates with GN-BSIs. If a correlation between gut colonisation and invasive infections is confirmed, 298
easy-to-collect rectal swab data could be used as a proxy, at the patients- or NICU-level, to inform 299
empirical antibiotic treatment in neonates with suspected BSIs. In the LMIC setting, blood cultures are 300
infrequently obtained from neonates, thus readily obtained rectal swabs could be used as a predictor 301
of MDR pattern at unit-level and help identify the optimal antibiotic regimens to be used. In HIC, 302
demonstrating a correlation between colonisation and invasive infections might help define the best 303
strategies for Infection Prevention and Control (e.g. cohorting babies during hospital outbreaks) and 304
to select babies who would benefit most from broad-spectrum antibiotics (for targeted clinical 305
management) and those who can receive more narrow-spectrum antibiotics. 306
15
TRANSPARENCY DECLARATION 307
Conflict of Interests: Mike Sharland reports other from Pfizer, GSK, outside the submitted work; and 308
Julia Bielicki declared that her husband is senior corporate counsel at Novartis International AG, Basel, 309
Switzerland and owns stock and stock options. 310
Funding: This study did not receive any direct funding. 311
Contributors statement: All authors contributed to the conception and design of the study. LF and CT 312
collected the data. LF, YH, JB, and MS contributed to the analysis of the data. All authors contributed 313
to the interpretation of the data. LF, CT, and JB wrote the first draft of the manuscript. All authors 314
revised the manuscript critically for important intellectual content. All authors approved the final 315
version of the manuscript to be submitted. 316
16
REFERENCES 317
1. Russell AB, Sharland M, and Heath PT. Improving antibiotic prescribing in neonatal units: time 318
to act. Arch Dis Child Fetal Neonatal Ed, 2012. 97(2): p. F141-6. 319
2. Le Doare K, Bielicki J, Heath PT, Sharland M. Systematic Review of Antibiotic Resistance Rates 320
Among Gram-Negative Bacteria in Children With Sepsis in Resource-Limited Countries. J 321
Pediatric Infect Dis Soc, 2015. 4(1): p. 11-20. 322
3. Stapleton PJ, Murphy M, McCallion N, Brennan M, Cunney R, Drew RJ. Outbreaks of extended 323
spectrum beta-lactamase-producing Enterobacteriaceae in neonatal intensive care units: a 324
systematic review. Arch Dis Child Fetal Neonatal Ed, 2016. 101(1): p. F72-8. 325
4. Almuneef MA, Baltimore RS, Farrel PA, Reagan-Cirincione P, Dembry LM. Molecular typing 326
demonstrating transmission of gram-negative rods in a neonatal intensive care unit in the 327
absence of a recognized epidemic. Clin Infect Dis, 2001. 32(2): p. 220-7. 328
5. Hartz LE, Bradshaw W, Brandon DH. Potential NICU Environmental Influences on the 329
Neonate's Microbiome: A Systematic Review. Adv Neonatal Care, 2015. 15(5): p. 324-35. 330
6. Singh N, Patel KM, Léger MM, et al. Risk of resistant infections with Enterobacteriaceae in 331
hospitalized neonates. Pediatr Infect Dis J, 2002. 21(11): p. 1029-33. 332
7. Basu S. Neonatal sepsis: the gut connection. Eur J Clin Microbiol Infect Dis, 2015. 34(2): p. 215-333
22. 334
8. Parm Ü, Metsvaht T, Sepp E, et al. Mucosal surveillance cultures in predicting Gram-negative 335
late-onset sepsis in neonatal intensive care units. J Hosp Infect, 2011. 78(4): p. 327-32. 336
9. Das P, Singh AK, Pal T, Dasgupta S, Ramamurthy T, Basu S. Colonization of the gut with Gram-337
negative bacilli, its association with neonatal sepsis and its clinical relevance in a developing 338
country. J Med Microbiol, 2011. 60(Pt 11): p. 1651-60. 339
10. McGinigle KL, Gourlay ML, Buchanan IB. The use of active surveillance cultures in adult 340
intensive care units to reduce methicillin-resistant Staphylococcus aureus-related morbidity, 341
mortality, and costs: a systematic review. Clin Infect Dis, 2008. 46(11): p. 1717-25. 342
17
11. Wells GA, Shea B, O'Connell D, Peterson J, Welch V, Losos M, Tugwell P. The Newcastle-Ottawa 343
Scale (NOS) for assessing the quality of nonrandomised studies in meta-analyses. Available 344
from: http://www.ohri.ca/programs/clinical_epidemiology/oxford.asp. (2005) Accessed 31 345
May 2017 346
12. Fitchett EJ, Seale AC, Vergnano S, et al. Strengthening the Reporting of Observational Studies 347
in Epidemiology for Newborn Infection (STROBE-NI): an extension of the STROBE statement 348
for neonatal infection research. Lancet Infect Dis, 2016. 16(10): p. e202-13. 349
13. Liberati A, Altman DG, Tetzlaff J, et al. The PRISMA statement for reporting systematic reviews 350
and meta-analyses of studies that evaluate health care interventions: explanation and 351
elaboration. J Clin Epidemiol, 2009; 62: e1-34. 352
14. Higgins JPT, Green S. Cochrane Handbook for Systematic Reviews of Interventions Version 353
5.1.0 The Cochrane Collaboration, 2011. Available from: www.cochrane-handbook.org. 354
(2011) Accessed 22 July 2017 355
15. Song F, Sheldon TA, Sutton AJ, Abrams KR, Jones DR. Methods for exploring heterogeneity in 356
meta-analysis. Eval Health Prof, 2001; 24(2): 126-51. 357
16. Freeman MF, Tukey JW. Transformation related to the angular and the square root. Ann Math 358
Statist, 1950; 21: 607-11. 359
17. Akturk H, Sutcu M2, Somer A, et al. Carbapenem-resistant Klebsiella pneumoniae colonization 360
in pediatric and neonatal intensive care units: Risk factors for progression to infection. Braz J 361
Infect Dis, 2016. 20(2): p. 134-140. 362
18. Biran V, Gaudin A, Mariani-Kurdjian P, Doit C, Bingen E, Aujard Y. [Implication of extended-363
spectrum beta-lactamase enterobacteriaceae in nosocomial infections in neonates]. Arch 364
Pediatr, 2010. 17 Suppl 4: p. S150-3. 365
19. Boo NY, Ng SF, Lim VK. A case-control study of risk factors associated with rectal colonization 366
of extended-spectrum beta-lactamase producing Klebsiella sp. in newborn infants. J Hosp 367
Infect, 2005. 61(1): p. 68-74. 368
18
20. Graham PL 3rd, Della-Latta P, Wu F, Zhou J, Saiman L. The gastrointestinal tract serves as the 369
reservoir for Gram-negative pathogens in very low birth weight infants. Pediatr Infect Dis J, 370
2007. 26(12): p. 1153-1156. 371
21. Gundes S, Arisoy AE, Kolayli F, et al. An outbreak of SHV-5 producing Klebsiella pneumoniae 372
in a neonatal intensive care unit; meropenem failed to avoid fecal colonization. New 373
Microbiol, 2005. 28(3): p. 231-6. 374
22. Mammina C, Di Carlo P, Cipolla D, et al. Nosocomial colonization due to imipenem-resistant 375
Pseudomonas aeruginosa epidemiologically linked to breast milk feeding in a neonatal 376
intensive care unit. Acta Pharmacol Sin, 2008. 29(12): p. 1486-92. 377
23. Mustapa, M, Egyepong J, Abdul-Rahman AK. Predictive value of admission surface swabs in 378
early-onset neonatal sepsis in extremely low birth weight (ELBW) infants in a neonatal 379
intensive care unit (NICU). [Abstract presented at the 5th Congress of the European Academy 380
of Paediatric Societies, EAPS 2014 Barcelona Spain]. Arch Dis Child, 2014. 99: p. A432-A433. 381
24. Oteo J, Cercenado E, Vindel A, et al. Outbreak of multidrug-resistant CTX-M-15-producing 382
Enterobacter cloacae in a neonatal intensive care unit. J Med Microbiol, 2013. 62(PART4): p. 383
571-575. 384
25. Pessoa-Silva CL, Meurer Moreira B, Câmara Almeida V, et al. Extended-spectrum beta-385
lactamase-producing Klebsiella pneumoniae in a neonatal intensive care unit: risk factors for 386
infection and colonization. J Hosp Infect, 2003. 53(3): p. 198-206. 387
26. Smith A, Saiman L, Zhou J, Della-Latta P, Jia H, Graham PL 3rd. Concordance of Gastrointestinal 388
Tract Colonization and Subsequent Bloodstream Infections With Gram-negative Bacilli in Very 389
Low Birth Weight Infants in the Neonatal Intensive Care Unit. Pediatr Infect Dis J, 2010. 29(9): 390
p. 831-5. 391
27. Suviste J, Gray J, Morgan I, Patel M. Rectal swabs: An increasingly important component of 392
nicu infection surveillance programmes? [Abstract presented at the 4th Congress of the 393
European Academy of Paediatric Societies Istanbul Turkey]. Arch Dis Child, 2012. 97: p. A334. 394
19
28. Velasco C, Rodríguez-Baño J, García L, et al. Eradication of an extensive outbreak in a neonatal 395
unit caused by two sequential Klebsiella pneumoniae clones harbouring related plasmids 396
encoding an extended-spectrum beta-lactamase. J Hosp Infect, 2009. 73(2): p. 157-163. 397
29. Cassettari VC, da Silveira IR, Dropa M, et al. Risk factors for colonisation of newborn infants 398
during an outbreak of extended-spectrum beta-lactamase-producing Klebsiella pneumoniae 399
in an intermediate-risk neonatal unit. J Hosp Infect, 2009. 71(4): p. 340-7. 400
30. Gbaguidi-Haore H, Talon D, Thouverez M, Menget A, Bertrand X. Molecular epidemiology of 401
Enterobacter cloacae in a neonatal department: a 2-year surveillance study. Eur J Clin 402
Microbiol Infect Dis, 2008. 27(8): p. 643-8. 403
31. Gupta A, Della-Latta P, Todd B, et al. Outbreak of extended-spectrum beta-lactamase-404
producing Klebsiella pneumoniae in a neonatal intensive care unit linked to artificial nails. 405
Infect Control Hosp Epidemiol, 2004. 25(3): p. 210-5. 406
32. Haase R, Worlitzsch D, Schmidt F, Kulka R, Kekulé AS, Körholz D. Colonization and infection 407
due to multi-resistant bacteria in neonates: a single center analysis. Klin Padiatr, 2014. 226(1): 408
p. 8-12. 409
33. Litzow JM, Gill CJ, Mantaring JB, et al. High frequency of multidrug-resistant gram-negative 410
rods in 2 neonatal intensive care units in the Philippines. Infect Control Hosp Epidemiol, 2009. 411
30(6): p. 543-9. 412
34. Macnow T, O'Toole D, DeLaMora P, et al. Utility of surveillance cultures for antimicrobial 413
resistant organisms in infants transferred to the neonatal intensive care unit. Pediatr Infect 414
Dis J, 2013. 32(12): p. e443-50. 415
35. Mammina C, Di Carlo P, Cipolla D, et al. Surveillance of multidrug-resistant gram-negative 416
bacilli in a neonatal intensive care unit: prominent role of cross transmission. Am J Infect 417
Control, 2007. 35(4): p. 222-30. 418
20
36. Parm U, Metsvaht T, Sepp E, et al. Risk factors associated with gut and nasopharyngeal 419
colonization by common Gram-negative species and yeasts in neonatal intensive care units 420
patients. Early Hum Dev, 2011. 87(6): p. 391-9. 421
37. Rettedal S, Høyland Löhr I, Natås O, Sundsfjord A, Øymar K. Risk factors for acquisition of CTX-422
M-15 extended-spectrum beta-lactamase-producing Klebsiella pneumoniae during an 423
outbreak in a neonatal intensive care unit in Norway. Scand J Infect Dis, 2013. 45(1): p. 54-8. 424
38. Richards C, Alonso-Echanove J, Caicedo Y, Jarvis WR. Klebsiella pneumoniae bloodstream 425
infections among neonates in a high-risk nursery in Cali, Colombia. Infect Control Hosp 426
Epidemiol, 2004. 25(3): p. 221-5. 427
39. Roy S, Viswanathan R, Singh A, Das P, Basu S. Gut colonization by multidrug-resistant and 428
carbapenem-resistant Acinetobacter baumannii in neonates. Eur J Clin Microbiol Infect Dis, 429
2010. 29(12): p. 1495-1500. 430
40. The World Bank list of economies. Available from: 431
http://siteresources.worldbank.org/DATASTATISTICS/Resources/CLASS.XLS (2016). Accessed 432
5 January 2017. 433
41. Versporten A, Bielicki J, Drapier N, Sharland M, Goossens H; ARPEC project group. The 434
Worldwide Antibiotic Resistance and Prescribing in European Children (ARPEC) point 435
prevalence survey: developing hospital-quality indicators of antibiotic prescribing for children. 436
J Antimicrob Chemother, 2016. 71(4): p. 1106-17. 437
42. Simon A, Tenenbaum T. Surveillance of multidrug-resistant Gram-negative pathogens in high-438
risk neonates--does it make a difference? Pediatr Infect Dis J, 2013. 32(4): p. 407-9. 439
43. Collado MC, Cernada M, Neu J, Pérez-Martínez G, Gormaz M, Vento M. Factors influencing 440
gastrointestinal tract and microbiota immune interaction in preterm infants. Pediatr Res, 441
2015. 77(6): p. 726-31. 442
21
44. Masse J, Elkalioubie A, Blazejewski C, et al. Colonization pressure as a risk factor of ICU-443
acquired multidrug resistant bacteria: a prospective observational study. Eur J Clin Microbiol 444
Infect Dis, 2016. Dec 20. doi: 10.1007/s10096-016-2863-x. 445
45. Vergnano S. Decolonization and decontamination: what's their role in infection control? Curr 446
Opin Infect Dis, 2015. 28(3): p. 207-14. 447
46. Birt J, Le Doare K, Kortsalioudaki C, Lawn J, Heath PT, Sharland M. Lack of evidence for the 448
efficacy of enhanced surveillance compared to other specific interventions to control neonatal 449
healthcare-associated infection outbreaks. Trans R Soc Trop Med Hyg, 2016. 110(2): p. 98-450
106. 451
47. Tacconelli E, Cataldo MA, Dancer SJ, et al. ESCMID guidelines for the management of the 452
infection control measures to reduce transmission of multidrug-resistant Gram-negative 453
bacteria in hospitalized patients. Clin Microbiol Infect, 2014. 20 Suppl 1: p. 1-55. 454
48. Tumbarello M, Trecarichi EM, Bassetti M, et al. Identifying patients harboring extended-455
spectrum-beta-lactamase-producing Enterobacteriaceae on hospital admission: derivation 456
and validation of a scoring system. Antimicrob Agents Chemother, 2011. 55(7): p. 3485-90. 457
22
Table 1: Meta-regression to determine the factors that account for the heterogeneity between studies in the baby-level
Variable Coefficient p-value 95%CI lower 95%CI upper
Screening timing 0.197 0.594 -3.193 3.589
GDPa classification -0.273 0.417 -2.939 2.392
During outbreak (Y/N) 0.275 0.412 -2.370 2.921
Pathogen -0.266 0.422 -2.910 2.378
aGDP: Gross domestic product
23
Table 2: Colonisation pressure and rate of Bloodstream Infections in studies included in the unit-level analysis
Author, year Population (n of
screened babies)
N of colonised babies Colonisation pressure (%) N of infected babies
(in the same period)
BSI rate (%)
Cassettari VC, 2009 [29] 120 27 22.5 7 5.8
Das P, 2011 [9] 242 198 81.8 32 13.2
Gbaguidi-Haore H, 2008 [30] 735 166 22.6 29 3.9
Gupta A, 2004 [31] 73 14 19.2 6 8.2
Haase R, 2014 [32] 635 27 4.3 4 0.6
Litzow JM, 2009 [33] 1,831 1,017 55.5 358 19.6
Macnow T, 2013 [34] 1,475 15 1.0 8 0.5
Mammina C, 2007 [35] 210 116 55.2 25 11.9
Parm U, 2011 [36] 276 154 55.8 27 9.8
Rettedal S, 2013 [37] 469 58 12.4 1 0.2
Richards C, 2004 [38] 69 8 11.6 0 0.0
Roy S, 2010 [39] 228 25 11.0 19 8.3
aBSI: Bloodstream infection
24
Figure 1
25
Figure 2
26
Figure legends
Figure 1: Flowchart and study selection
Figure 2: Random effects meta-analysis for estimated risk ratio at the baby-level by groups
(Abbreviations: CI, confident interval; RR, risk ratio; HIC, high income country; UMIC, upper middle income
country)
Figure 1S: Random effect meta-analysis for infection rate at the unit-level by groups